Measurements of the subjective visual horizontal (SVH) and the subjective visual eye level (SVEL) were made in a large swing-out gondola centrifuge. Rotation of the centrifuge was anti-clockwise, as seen from above. Test subjects were seated upright in the gondola, facing forwards. In front of the subject there was either a luminous line, which could be rotated about the visual axis, or a laser dot, which could be moved up or down. Thirteen normal subjects were asked to indicate, by adjusting the line or dot, what they perceived as horizontal in roll (SVH) and pitch (SVEL). Measurements were made in lg environment and during 10 min at 2G. Initially after acceleration of the centrifuge the SVH was tilted relative to the gravitoinertial horizontal. The direction of this tilt was dependent on the subject´s strategy for spatial orientation, which could be either allocentric (referring to the surroundings) or egocentric (referring to the co-ordinate system of the own body). The interindividual variability of the initial SVH-tilt was large, ranging from a few degrees to more than 40 degrees. The magnitude of tilt was larger for the allocentric strategy (mean 19 degrees) than for egocentric strategy (10 degrees). The time constant for exponential decay was 2-3 min. At 2G the SVEL was lower than in lg environment. The deviation from the gravitoinertial eye level gradually increased, approaching an asymptotic maximum with a time constant of 2-3 min. The asymptote was dependent on the strategy of the subject; the mean values were 25 degrees (allocentric strategy) and 13 degrees (egocentric strategy). In conclusion, the normal subjects showed a large interindividual variability, which may be related to several factors, such as mental strategy, relative dependency on otolith versus semicircular canal information on changes in head orientation, and the memory for semicircular canal information on changes in static head position. The results are viewed in the context of spatial disorientation.